Hey everybody, I learned the hard way recently that many (most?) linear voltage regulators have a reverse leakage path equivalent to a diode drop. In case I've phrased that wrong, what I mean is that if you leave them input floating, and the output has voltage (like when you disconnect the power source at the input, but there are still caps holding charge on the output,) current will flow backwards through the regulator pulling the input pin up to roughly a diode drop below the output.

What I'm wondering now is if this is also generally true of buck converters. This question came up because I've been trying to help in another thread here (https://forum.allaboutcircuits.com/...rown-out-condition.150235/page-2#post-1287007) and I was concerned about reverse leakage through a regulator, only realizing pretty late in the thread that it was a buck converter, not a linear regulator.

Google has failed me hard on this one - every search result pertains to polarity reversal, connecting the Vin and ground backwards. What I'm interested in is what happens when Vout is greater than Vin.

I've also tried looking at a few datasheets with no luck. I don't see any spec relating to these conditions. Am I overlooking something? Here's one example datasheet:

https://www.diodes.com/assets/Datasheets/AP1501.pdf

 

It is difficult to say with an integrated controller/power IC without knowing the specifics of the internal circuit. If the switch is a bipolar transistor, the path may not exist,. When the schematic of the internals is a block diagram instead of a full schematic with all the little details, it may be impossible to detect a path that is there. For the Diodes part, the path would appear not to exist, but in a critical application I would either consult the manufacturer or make some measurements.

In a buck converter with a discrete FET the body diode of the FET makes a path between output and input. This can be an issue with something like a battery charger where the battery remains connected when the input supply is lost. It can keep the control circuitry up and running and could result in a battery discharge current of several milliamps.

The reverse path can sometimes be via parts not intended to carry any significant current. In situations where such a path exists and there is a possibility of the input supply dropping abruptly and presenting a low impedance to the current through the device, it is usually wise to add a discrete diode in reverse between input an output to create a well-defined path that can handle discharge of the output capacitors.

If you need to block reverse current, then you probably need to add a diode, either conventional or "ideal", with the consequent added complexity and reduction of efficiency.

 

It is difficult to say with an integrated controller/power IC without knowing the specifics of the internal circuit. If the switch is a bipolar transistor, the path may not exist,. When the schematic of the internals is a block diagram instead of a full schematic with all the little details, it may be impossible to detect a path that is there. For the Diodes part, the path would appear not to exist, but in a critical application I would either consult the manufacturer or make some measurements.

In a buck converter with a discrete FET the body diode of the FET makes a path between output and input. This can be an issue with something like a battery charger where the battery remains connected when the input supply is lost. It can keep the control circuitry up and running and could result in a battery discharge current of several milliamps.

The reverse path can sometimes be via parts not intended to carry any significant current. In situations where such a path exists and there is a possibility of the input supply dropping abruptly and presenting a low impedance to the current through the device, it is usually wise to add a discrete diode in reverse between input an output to create a well-defined path that can handle discharge of the output capacitors.

If you need to block reverse current, then you probably need to add a diode, either conventional or "ideal", with the consequent added complexity and reduction of efficiency.

Thanks! As usual, a clear and helpful explanation.

Given that there's no general rule per se, I guess I'll always start from the assumption that there *could* be such a path, then analyze its possible side effects to see whether I need more detailed part research and/or additional components for protection.